Computational Models Ecosystems

The future of digital technologies for crops

Cyber physical technologies such as earth observation and biophysical crop modelling can play a major role in production, sustainability, and profitability of production systems in the face of climate change.

When should I begin applying variable-rate fertilizer and where? Do my south fields have a pest or disease problem?

Will planting delays due to heavy rains reduce regional crop yields? What about the hot and dry summer due to El Niño? Which fields need to be hayed-off due to crop stresses like frost, heat, or drought?  

Digital technologies can be used to mitigate and predict production losses with real-time data. They can be used for management interventions such as fertilization, pest control, stress detection, irrigation management, and weed control before they negatively impact crop productivity. Accurate information on the spatial distribution and growth dynamics of cropping is essential for assessing potential risks to food security and are also critical for evaluating the market trends at regional, national and even global levels.

Dr Andries Potgieter, Associate Professor at The University of Queensland research fellow at the Queensland Alliance for Agriculture and Food Innovation is the lead author of a new paper published in in silico Plants that reviews what’s new and what’s next in the world of digital technologies for crops. According to Potgieter, “This paper, highlights the advances made in earth observation, machine learning and cloud computing technologies, specifically during the last 5-years. It furthermore, discusses the fusion of such technologies with targeted knowledge-based biophysical systems which will lead to more cutting-edge applications within agriculture. Such integrated predictive physical systems have the potential to mitigate the impact of climate extremes and change on agricultural production systems. Specifically, in Australia where the projected impacts of future climates on food production is of more concern to the sustainability and resilience of industries to cope with it.”

Proposed set-up of a multiple digital sensor platform to record information at a detailed crop validation site in recently funded GRDC project (CropPhen). This targets the development of new approaches for the accurate monitoring of crop phenology and discriminating of crop types (copyright QAAFI).

The authors first review remote- and proximal-sensing technologies. Currently, there are more than 140 earth observation (EO) satellites in orbit, carrying sensors that measure visible, infrared and microwave regions of the electromagnetic spectrum of terrestrial vegetation. Remote sensing allows crops across large areas to be monitored over time and with precise repeated data. This data can be used (among other things) to discriminate between different crop types (e.g., wheat, barley, chickpeas and canola) and quantify cropping area at regional and landscape levels.

Recent advances in sensor technologies have led to the rapid increase in the use of drones or unmanned aerial vehicles (UAV) carrying proximal sensors. These sensors are deployed at the field level and have higher spatial and temporal resolution than remote EO sensors. This data can be used to monitor crop type, crop canopy and to quantify canopy structural and biophysical parameters (e.g., leaf area index, transpiration, photosynthetic pigments). Information from both platforms can be integrated for finer scale crop classifications.

The paper presents studies using remote sensing and UAVs along with the crops being assessed, sensors and algorithms. It also describes the pros and cons, as well as the applications for the sensor types and sensor platforms commonly used with agriculture.

After the data is collected, algorithms are used to classify the remote sensing data. Traditionally, knowledge-based approaches are explicitly developed by experts using theory-based equations. The advance in computing power (e.g., cloud computing and high-resolution imagery) has led to the development of machine learning and more complex deep learning approaches. Machine learning builds models directly from data without relying on predetermined equations. The strengths of machine learning include the capacity to handle data of high dimensionality and to map classes with very complex characteristics. The authors give specific examples of various machine learning and deep learning techniques.

Crop models deliver the predictive power of remote sensing data. Crop models allow predictions of yield and phenological development stages for both current and/or projected climate scenarios. Combining real-time RS data with crop models can provide real-time information on crop growth and development and predict phenology and yield. The resulting tools are more important than ever as climate variability and change have an increasing influence on crop production and food security.

Information services based on remote sensing data are now available to the public via commercial platforms. The paper provides examples of these platforms and the information they provide.

What is the future of digital technologies in agriculture?

Crop growth and development are mainly a function of the interactions between genotype, environment, and management (G × E × M). Thus, these interactions are the key to realizing further gains in global crop yields and ensuring future food security. The environment cannot be changed but different cultivars (genotypes) can be planted with adjusting management practices (e.g. sowing dates, row spacings, seed density) tailored to the environment and future climate.

Farmers know that not all fields are equal: some always produce more, others always less, while other fields vary in their production capacity from one year to the next. “Planning for an average season does not make sense, since no rainfall season is the same across a farm in any given year”, Dr Potgieter said. Furthermore, the ability to make informed decisions from regional yield estimates are limited by this variability in G × E × M at a local scale.

Up to now, technologies and integrative approaches were limited in effectively harnessing all dimensions of the targeted data available to achieve higher accuracies in crop phenology or crop-type estimation. A/Prof Potgieter is currently the lead on a national project (supported by the Grains Research and Development Corporation and The University of Queensland) exploring new frontiers in the application and/or developing novel integrative metrics that can effectively help solve this complex issue. Specifically, we aim to harness all dimensions, i.e. temporal, spatial, spectral, and physiological, of the targeted data available to achieve higher accuracies in crop phenology or crop-type estimation. “This will result in integrated solution to allow accurate development, validation, and scalability of predictive tools for crop phenology mapping (G) at within-field scales (M), across extensive cropping areas (E). Working closely with industry partners in the digital space it is anticipated that producers will have access to these digital tools that will aid them in making more informed decisions, targeted to the G × E × M for each field, early on. Thus, reducing input costs, optimizing risk management and improve whole-farm profitability” says Dr Potgieter.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.